OpenAI disbands preparedness team in July 2026, shifting safety approach

Technology Aug 17, 2026

OpenAI has disbanded its Preparedness team, the group that used to watch out for the biggest dangers from its AI models.

This happened in July 2026, not long after another safety-focused team was also dissolved.

The move marks a shift away from OpenAI's earlier research-heavy approach to handling AI risks.

Dylan Scandinaro targets self-improving AI risks

After the shake-up, senior staff were moved to existing OpenAI teams tackling particular areas of preparedness such as cyber and bio.

Dylan Scandinaro, who led the Preparedness team, is now zeroing in on risks from self-improving AI.

All this comes as OpenAI faces leadership exits and gears up for a possible $1 trillion IPO, while rival Anthropic is quickly gaining ground in the race for smarter (and safer) AI.